Hello, Tejun. 在 2026/9/22 03:37, Tejun Heo 写道: > Hello, Tao. > > On Fri, 18 Sep 2026 13:49:56 +0800, Tao Cui wrote: > >> The registration and binding model follows the TCP congestion >> control model registration pattern: registering a struct_ops makes >> the model available by its name, while io.cost.model binds one >> registered model to a device with "model=<name>" and restores the >> builtin model with "model=linear". > > Alexei is right, this isn't what I meant. One struct_ops instance per > device: attaching binds the instance to the device and switches the device > to the model, detaching switches it back to the builtin. That removes the > name registry, the lifecycle list and the refcounting on top of the > struct_ops map reference. Putting the device id in the struct_ops the way > hid_bpf_ops does works fine. I don't have a strong opinion on how the > device is identified. >
I misunderstood what you meant by per-device. v7 reworks this in the hid_bpf_ops style: one struct_ops instance per device, with the target device identified through the struct_ops instance. .reg binds it to the device and .unreg detaches it and restores the builtin model. The name registry, lifecycle list and additional refcounting layer are gone; attaching a second model to an already-bound device fails with -EBUSY. >> u64 calc_cost(u64 opf, u64 nbytes, sector_t sector, >> struct blkcg *blkcg, u64 model_flags) > > Why not pass the bio itself? The model can read whatever it needs from it. > The merge indicator still needs the flags argument as it isn't a property > of the bio. > calc_cost() takes the bio itself now, so the model can read the operation, size, sector and issuing cgroup from it; only the merge indicator stays in the separate flags argument. >> blkcg is passed so the model can key >> per-cgroup state; state stored in BPF_MAP_TYPE_CGRP_STORAGE >> follows the cgroup lifetime, and optional blkcg_online()/ >> blkcg_offline() callbacks mirror the css lifecycle for models >> which want eager setup or teardown. > > With a per-device instance, these should be bound to the iocg pd init and > free rather than the blkcg css. That gives the model the same per cgroup > per device lifetime as the builtin cursor and drops the global mutex from > the cgroup online and offline paths. > Done: the callbacks are bound to the iocg pd init and free paths, giving the model the same per-(cgroup, device) lifetime as the builtin cursor and avoiding the global mutex in the cgroup online/offline paths. >> The completion-time request sizing for the >> latency met/missed accounting still uses the builtin coefficients >> (the request's bio, and with it the issuing cgroup, is gone by then); >> extending the model there is left open by this interface. > > Let's be explicit here. For now, the struct_ops should carry the transfer > cost coefficients, vtime per page for reads and writes, and the builtin > latency tracking and vrate adjustment should use these instead of the > linear coefficients while the model is attached. > The pricing paths now use the transfer cost coefficients (vtime per page for reads and writes) provided by the struct_ops while the model is attached. This covers completion-time request sizing, latency tracking and vrate adjustment, which no longer use the linear coefficients in this case. There is one remaining model call in the completion path in my current tree which does not belong to this interface; it will be removed in the posted version. > Down the line, a model should be able to take over the QoS side too. The > vrate adjustment is driven by an empirically derived table and being able > to replace it is a good part of the value of a pluggable model. The shape > I have in mind is an rq_done(struct request *) callback which disables the > builtin latency tracking and periodic vrate adjustment, with the model > modulating directly through kfuncs to set the base vrate and the > saturation state and to read the block layer clock. That doesn't have to > be part of this series. > Understood. The QoS-side takeover (rq_done() and kfunc-based vrate and saturation control) stays out of this series. > While a model is attached, the linear coefficients are inert. Coefficient > writes can be stored and take effect again on detach, autop shouldn't step > profiles based on vrate the same way it doesn't when the cost model is > user-set, and io.cost.model can report the model in the model field, in a > form which identifies it as a BPF model, with ctrl still describing the > coefficients. > Yes, coefficient writes are stored and take effect again on detach. autop will only step profiles when no model is attached, matching the user-set cost model behavior. io.cost.model readback will report the BPF model in the model field while ctrl continues to describe the coefficient source (user/auto). I'll send the reworked series. Thanks, Tao > Thanks. > > -- > tejun
